This article traces Monte Zweben's journey in building expert systems and bringing them into the business world. He started out learning about rule-based systems at Carnegie Mellon in the 1970s and implementing constraint-based systems at NASA and MITRE in the 1980s. He built upon this early experience to co-found Red Pepper Software, a company that specialized in planning and scheduling manufacturing tasks. Following Red Pepper's success, he founded his second startup, Blue Martini, which used machine learning to enhance e-commerce businesses. Since the arrival of Big Data in the 21st century, he has watched the industry transform and was a part of the data-driven marketing revolution at Rocket Fuel. Throughout his career, his work has centered on expert systems and how they can be used to help people. This is the story of his journey and the evolution of AI in business.

MESSAGES DON'T WORK if customers p aren't receptive to them. To boost the odds of being J greeted by eager eyes and hungry ears, marketers ask questions that target their audiences. Who should receive the message? What should the content of the message be? How should we deliver the message? The one question they rarely ask is, when should we deliver the message? Yet in marketing, timing is arguably the most important variable of all. Promotions are like weather reports and news bulletins-people need them when they need them. Too early and they are forgotten. Too late and they are ignored. Most people don't want to hear from most companies most of the time, and in an era of marketing overloadcharacterized by irrelevance as well as volume unsolicited communication can provoke apathy or worse, resentment. The business press has lately described a "chaos scenario" in which traditional media (television, print) and established formats (the 30-second spot) are in decline but new, more effective media and formats are still evolving. In the messy interim, advertisers find their

Marketers planning promotional campaigns ask questions to boost the odds that the messages will be accepted: Who should receive each message? What should be its content? How should we deliver it? The one question they rarely ask is, when should we deliver it? That's too bad, because in marketing, timing is arguably the most important variable of all. Indeed, there are moments in a customer's relationship with a business when she wants to communicate with that business because something has changed. If the company contacts her with the right message in the right format at the right time, there's a good chance of a warm reception. The question of "when" can be answered by a new computer-based model called "dialogue marketing," which is, to date, the highest rung on an evolutionary ladder that ascends from database marketing to relationship marketing to one-to-one marketing. Its principle advantages over older approaches are that it is completely interactive, exploits many communication channels, and is "relationship aware": that is, it continuously tracks every nuance of the customer's interaction with the business. Thus, dialogue marketing responds to each transition in that relationship at the moment the customer requires attention. Turning a traditional marketing strategy into a dialogue-marketing program is a straightforward matter. Begin by identifying the batch communications you make with customers, then ask yourself what events could trigger those communications to make them more timely. Add a question or call to action to each message and prepare a different treatment or response for each possible answer. Finally, create a series of increasingly urgent calls to action that kick in if the question or call to action goes unanswered by the customer. As dialogue marketing proliferates, it may provide the solid new footing that Madison Avenue seeks.

Los expertos en marketing que planifican campanas promocionales hacen preguntas para aumentar las posibilidades de que sus mensajes seran aceptados: ?Quien deberia recibir los mensajes? ?Cual deberia ser su contenido? ?Como deberiamos entregarlos? La unica pregunta que rara vez se hace es, ?cuando deberiamos entregarlos?. Eso esta muy mal, porque se puede decir que en marketing la eleccion del momento es la variable mas importante de todas. En efecto, existen momentos en la relacion de una clienta con una empresa, en los que quiere comunicarse con esa empresa porque algo ha cambiado. Si la compania la contacta con el mensaje correcto, con el formato adecuado, en el momento preciso, existe una buena posibilidad de una recepcion calida. La pregunta del ?cuando? puede ser respondida por un nuevo modelo basado en la computacion, llamado ?marketing de dialogo, el cual es el peldano mas alto en la escala evolutiva que asciende desde el marketing de base de datos hasta el marketing uno a uno. Su principal ventaja sobre los enfoques mas antiguos es que es completamente interactivo, aprovecha muchos canales de comunicacion y es ?consciente de las relaciones?: esto es, rastrea todos los matices de la interaccion de los clientes con la empresa. De este modo, el marketing de dialogo responde a cada transicion en esa relacion, en el momento en que el cliente requiere atencion. Convertir una estrategia de marketing convencional en un programa de marketing de dialogo es un asunto directo. Comience por identificar las comunicaciones en lotes que usted tiene con los clientes, luego preguntese por los acontecimientos que las activarian para que sean mas oportunas. Luego haga una pregunta o un llamado a la accion a cada mensaje y prepare un trato o respuesta diferentes para cada respuesta posible. Finalmente, cree una serie de llamados a la accion crecientemente urgentes, que se activan si el llamado a la accion permanece sin ser respondido por el cliente.

Report describes continuing development of software for constraint-based scheduling system implemented eventually on massively parallel computer. Based on machine learning as means of improving scheduling. Designed to learn when to change search strategy by analyzing search progress and learning general conditions under which resource bottleneck occurs.